We have data generated for a large number of animals using the 16s Illumina MiSeq platform. We have been using the output from the 16s Metagenomics App from BaseSpace, which states that it provides "ultra-fast taxonomic classification of the bacterial 16s rRNA gene without the need for upfront OTU clustering." The reads are classified against the GreenGenes database, with species-level sensitivity according to Illumina descriptions.
We are doing this on a species whose gut microbiome has not been studied/published yet. I am wondering if we are taking the correct approach (by using the app's taxonomic classifications) as opposed to doing something by ourselves using QIIME or something similar?
Your thoughts and suggestions are appreciated.
U.
I'll admit I don't know much about BaseSpace, but what classification algorithm they use? Do they provide any documentation of the accuracy of their pipeline, validated against some control data? I don't think anyone believes you can reliably get species-level identification from 16S rRNA amplicons.
Personally I'd recommend you analyse the data yourself in whatever software package you're comfortable with. The methods are all well tested and standardised, and if you're going to end up using another pipeline for your OTU clustering/analysis anyway it will be better to perform your classification at the appropriate stage of analysis (for example, after quality-filtering, chimera checking etc) rather than before all the screening steps are applied to your data. I think your concern is correct, in that if you're studying a system which has never been sequenced before you'll want some well-accepted methods for your initial analysis (not that there's necessarily anything wrong with Illumina's method, but it's certainly not one I've seen cited anywhere).
For what it's worth, most of the common classification approaches these days are very fast anyway (the Bayesian/RDP method, SortMeRNA, usearch etc) so classification isn't really a step that will hold up your analysis.
Hi, as far as I can say, QIIME is a very good pipeline for analyzing NGS data. I would also recommend mothur. Mothur is much more userfriendly for non-bioinformatic people or beginners, although it might be tricky at soem point due to large datasets generated from Illumina. From my point of view, QIIME would be the first choice.
http://www.mothur.org/
I agree with David that cleaning/screening steps (such as chimera removal, quality filtering etc.) should be performed before clustering and classification. In addition, if you want to somehow "test" your method, you can consider following two different pipelines and comparing your results afterwards. For example, with my first set of 16S Illumina Miseq data, I tried to follow both the QIIME pipeline and a customized pipeline provided by a colleague based on usearch, and since the results were similar I then decided to proceed with QIIME. It works as a kind of control. Good luck with your analysis!
I am not familiar with BaseSpace App, but I recommend using QIIME. It is a validated pipeline and provides large number of options and methods in the analysis. As per the requirement, one can customize almost everything in QIIME.
BaseSpace app is using Kraken classifier and runs the miniKraken database. Just for the sake of information.
As a matter of fact, it's doing a decent job for a rapid check on what's in your samples. It's most probably not going to do the trick for a publication as others already mentioned
[not that it'd drastically change conclusions but...]
- Badges
- Science topic
- Similar topics
- Biological Science
- Microbiology
- Microorganisms